|
FairVote Analysis of
Kathy Dopp's 2004 Florida Analysis
November 2004
Kathy Dopp originally compared the 2004 presidential margins (Bush minus Kerry divided by total votes) in Florida counties with the 2004 party registration margins in those counties. Doing so, she found a correlation coefficient of .616 in Touchscreen/DRE counties, but only .009 in counties using Optical Scan Equipment. Hence, she speculated that the optical scans might have been the cause of election fraud. We hypothesized that party registration was a bad predictor, and instead compared the 2004 presidential margins with the 2000 presidential margins. This would give us a comparison of 2004 with actual voting patterns, rather than speculation. Our analysis showed a very high correlation coefficient between the 2004 and 2000 results, regardless of what type of machines the counties used, thereby debunking Dopp's hypothesis. Note that in 2000, Touchscreen/DRE machines were not being used. Our Optical Scan correlation coefficient was .970, while our Touchscreen/DRE coefficient was .978.
Dopp's scatterplot of her analysis (copied above) was followed with the following comments: "If there is a strong relation between voting and registration, then we expect most data points in the right-top (Republican Registration and Voting) and the left-bottom (Democratic Registration and Voting) quadrants. This is the case for the E-Touchscreen voting machines. If a voting machine produces random errors, then we expect data points in the left-top (Democratic Registration and Republican Voting) and the right bottom (Republican Registration and Democratic Voting) quadrants. We see a lot of data points in the left-top quadrant, but not in the right-bottom quadrant. This seems to suggest a biased error towards Republican vote counting in counties with strong Democratic Party Registration." We redrew the scatterplots, replacing party registration as a predictor with the 2000 results. Note that the graphical anomalies she highlighted now disappear.
To view the spreadsheet analysis of
our 2004 and 2000 Florida comparisons, click here.
|
